pc: Yield of a chemical process depending on the amount of a...

In schaarschmidt/statintcov: Multiple comparisons of treatments in presence of treatment-covariate interaction

Description

A simulated data set (similar to a data set presented by Milliken and Johnson, 2002), in which an interaction of a treatment effect with the quadratic term of a covariate is plausible.

Usage

data("pc")

Format

A data frame with 36 observations on the following 3 variables.

yield

a numeric vector, observed yield of a chemical process, the response variable of interest

x

a numeric vector, substance on which the chemical porocess depends, a potential covariate

additive

a factor with levels Control S1 S2, reprenting the treatments of interest: the process run without additive (control), and the process run in presence of two different additives

Details

Milliken and Johnson (2002) discuss the analysis of this data set using ANCOVA. It is obvious, that a quadratic term can be used to model the yields dependency on the covariante x. Milliken and Johnson show that either an interaction of the additive with x and a common quadratic term meay be used, or a common linear term and an interaction of the additive with a quadratic term for x may be used. The latter model is used here to illustrate the comparison of treatments in presence of treatment- covariate interaction, when interaction is with the quadratic term.

Source

The data set is simulated to resemble a data set presented in Milliken and Johnson, 2002: Table 4.9, page 108; model: Table 4.11, 4.13, page 110.

Examples

data("pc")

fitpc<-lm(yield ~ x + I(x^2) + additive + additive:I(x^2), data=pc)
anova(fitpc)

if(require("ggplot2")){
ggplot(pc, aes(y=yield, x=x)) + geom_point() + facet_wrap(~additive) 
}

schaarschmidt/statintcov documentation built on May 29, 2019, 3:26 p.m.